Quicktime MJPEG Question

OtakuMegane · Post by **OtakuMegane** » Thu Apr 22, 2004 3:02 pm

I've noticed for some time now, that using Quicktime's MJPEG compressor, 99% quality hits the 3MB-4MB/s range, and gradually decreases with lower quality. But putting it up one tiny notch to 100%(Best) the size suddenly jumps to 7.5MB-9MB/s. Is this just something caused by the particular equations used, or does Quicktime actually have a 'lossless' setting for JPEG?

AbsoluteDestiny · Post by **AbsoluteDestiny** » Thu Apr 22, 2004 6:11 pm

There is a thing called lossless mjpeg but it's usually a different way of compressing.

AFAIK the highest quality level onmost mjpeg codecs is twice as big as the next level down, much like mpeg at quantiser 1.

Tab. · Post by **Tab.** » Thu Apr 22, 2004 6:34 pm

Each quality implementation in JPEG is independent -- all are only used as different ways of scaling the default quantization matrix. I have no clue what Quicktime uses, but most implementations follow the IJG's example with the equation:

quality > 50: scale = 2 - (quality / 50)
quality <= 50: scale = 50 / quality

Larger scales quantize more.

However, as you might notice, that equation yields a 0 for quality 100. *IF* apple uses that equation, there's most likely an adjustment in post that sets all matrix values rounded to (or equal to) 0 to 1, giving the closest to lossless lossy can get; no quantization, only roundoff.

Of course, that's all only semi-useful, since the effect on filesize at that quality is unpredictable and varies with the content.

OtakuMegane · Post by **OtakuMegane** » Mon Apr 26, 2004 2:33 pm

Tab. wrote:Each quality implementation in JPEG is independent -- all are only used as different ways of scaling the default quantization matrix. I have no clue what Quicktime uses, but most implementations follow the IJG's example with the equation:
quality > 50: scale = 2 - (quality / 50)
quality <= 50: scale = 50 / quality
Larger scales quantize more.

However, as you might notice, that equation yields a 0 for quality 100. *IF* apple uses that equation, there's most likely an adjustment in post that sets all matrix values rounded to (or equal to) 0 to 1, giving the closest to lossless lossy can get; no quantization, only roundoff.

That's the type of thing I figured it probably was. Tthe rounding you mentioned is likely what's used by Quicktime. I currently just work at 99 quality, since I don't have enough disk space yet for more. Once I have enough, probably try 100, since leaving some more data in might help with final compression quality(I would assume so at least).